Hello @NaniOnyx07, I took a look at both internal and external documentation and here is what I found.
First, credit to @ThomazNeto for the thorough first pass. I checked his facts against the docs and he is spot on: environment version 5 shipped in February 2026, is supported through February 2029, and has no known-issues entry. His steps (try version 4 or 6, try a brand new empty notebook, check your Free Edition quota) are where I'd start too, so I'll build on them rather than repeat them.
One thing to understand about the banner. "Databricks is experiencing heavy load" is a generic workspace-level message, not a serverless error and not a version 5 error. In my experience it shows up when calls from your browser to the control plane are slow or getting throttled. It's telling you the workspace UI is struggling, and the environment version is just where you happened to be when it did. So there's likely nothing to fix in your version 5 setup.
Two questions that help:
Which cloud and region? The page Thomaz checked (status.databricks.com) covers AWS. Azure has its own at status.azuredatabricks.net, and GCP has a separate page as well. "Since yesterday evening" is a long time for a capacity blip, so if there's a regional issue on your cloud, it'll be there.
Is anyone else in the same workspace seeing it? If a colleague can attach fine, it's your session or your notebook. If nobody can, it's the workspace or the region.
Then a short sequence:
- Reset the browser side. Sign out, hard refresh, and try an incognito window or a different browser with extensions off. That banner can get sticky in a stale session.
- Check whether serverless itself is healthy in your workspace. Run a query on a serverless SQL warehouse, or create a one-cell job on serverless compute. If those work, capacity is fine and the problem is in the notebook attach path.
- Look at the notebook's environment. Open the Environment panel on the right, switch the version, click Apply, and attach again. If the notebook was imported or came in through Git as a source-format file, pay extra attention here: source format only carries code, so the environment settings may not have come with it. Credit to @Thomas_W and @LonaOsmani in an older thread (linked below) for spotting that imported notebooks with a stale pinned version wouldn't attach until they switched it. Also check the Dependencies list. Installing PySpark, or any library that pulls PySpark in, stops the session per the docs. If you see anything like that, remove it, then click the arrow next to Apply and choose Reset to defaults to clear the cached environment.
If you're on Free Edition, keep in mind that exceeding the quota shuts compute down for the rest of the day (data and settings are kept, and it comes back when the limit resets), and Free Edition isn't covered by support, so this thread is the right place. If you're on a paid workspace and it's still happening after the browser reset and a colleague sees it too, open a support case with the workspace ID, cloud and region, UTC timestamps, environment version, notebook URL, a screenshot, and any request or error ID shown. Keep those identifiers out of the public thread.
References:
Regards, Louis.
